Parrot 3.1.0 Released | parrot.org | Log: irclog.perlgeek.de/parrot/today | Goals: Get GSoC ideas on wiki. Merge/review tewk/select. Evaluate attribute_defs, rip out in branch
Set by moderator on 1 March 2011.
00:15 bubaflub left
kid51 ~~ 01:28
plobsing hi kid51 01:41
kid51 plobsing: Did you see trac.parrot.org/parrot/ticket/2037? 01:45
Or: trac.parrot.org/parrot/ticket/540 ?
plobsing kid51: yeah. that looks at least partially GC related. infnan had *nothing* to do with anything GCable.
kid51 I didn't think it did. 01:46
I looked at the final diff and saw nothing mysterious.
The fact that something subsequently "corrected" the problem is as scary as the problem itself.
plobsing possibly a makefile failure with old object files? 01:47
which would explain a rebuild fix
kid51 Doubtful; when I was bisecting, I included a 'make realclean; perl Configure.pl' every time.
Unless there were object files named in one version of the 'clean' target but not another. 01:48
But that's a remote chance.
plobsing possibly an installed parrot conflicting? 01:49
whiteknight rakudo build takes sooooooo long 01:50
src/gen/core.pm is huge 01:53
13.5 minutes total 01:55
02:10 kid51 is now known as kid51_at_dinner 02:31 whiteknight left 02:39 ingy joined
cotto ~~ 02:46
02:56 kid51_at_dinner is now known as kid51
bacek aloha, humans 03:09
kid51 bacek, Did you see trac.parrot.org/parrot/ticket/2037 ? 03:14
bacek kid51, yes. 03:15
I'll try to look at it later today
kid51 Thx
03:26 contingencyplan left
bacek kid51, how much memory do you have on this box? 03:30
kid51 256M 03:44
(my 7-y-o iBook G4)
bacek kid51, ok. 03:48
kid51 It's the same machine that has been our reference point for "small resources" for other GC-related questions. 03:49
And all of your GC stuff that's gone into master was tested on it first 03:50
dalek rrot: b3bae9c | bacek++ | / (4 files):
Remove interp->dynamic_env. Leftover of #876.
 03:55
bacek kid51, can you produce backtrace of segfault? 03:56
kid51 What would be the steps to do that? 03:57
bacek gdb --args ./parrot-nqp --target=pir --output=compilers/opsc/gen/Ops/Compiler/Grammar.pir compilers/opsc/src/Ops/Compiler/Grammar.pm 03:59
r
bt 10
03:59 bubaflub joined
kid51 It's late here, and I'll have to checkout that commit. Will try to get to it tomorrow. Thx. 04:01
bacek kid51, no worries. 04:05
you can do it on HEAD btw 04:06
kid51 But at HEAD I am not getting that failure 04:08
Will that suffice?
Problem: 04:10
On Darwin, we recently made a change to guarantee that dynamic linking will work correctly on an installed parrot 04:11
But the price we paid for that is that during dev/testing, while dynamic linking works from 'make', it doesn't work (as well) from, e.g., 'prove'. 04:12
To run 'prove' on many of the t/pmc/*.t tests now, I first have to say:
export DYLD_LIBRARY_PATH=/Users/jimk/gitwork/parrot/blib/lib 04:13
This is the first time I've tried to use 'gdb' on this machine since that change.
gdb --args ./parrot-nqp --target=pir --output=compilers/opsc/gen/Ops/Compiler/Grammar.pir compilers/opsc/src/Ops/Compiler/Grammar.pm
I get:
warning: Unable to read symbols for "/usr/local/lib/libparrot.dylib" (file not found). 04:14
Reading symbols for shared libraries ........ done
(gdb) r 04:15
Starting program: /Users/jimk/gitwork/parrot/parrot-nqp --target=pir --output=compilers/opsc/gen/Ops/Compiler/Grammar.pir compilers/opsc/src/Ops/Compiler/Grammar.pm
dyld: Library not loaded: /usr/local/lib/libparrot.dylib
Referenced from: /Users/jimk/gitwork/parrot/parrot-nqp
Reason: image not found
Program received signal SIGTRAP, Trace/breakpoint trap.
0x8fe05fc4 in __dyld__ZN4dyld5_mainEPK11mach_headeriPPKcS5_S5_ ()
nopaste "kid51" at 192.168.1.3 pasted "backtrace on Ops/Compiler/Grammar.pir problem" (16 lines) at nopaste.snit.ch/35586 04:16
04:20 bubaflub left 04:25 bubaflub joined, bubaflub left
bacek kid51, not really helpful... 04:31
kid51 I suspected as much. 04:32
plobsing try setting the env var for the launching GDB process
kid51 How would I do that?
plobsing just put it in front of the invocation 04:33
A=B gdb ...
kid51 I already have this in my .bashrc:
export DYLD_LIBRARY_PATH=/Users/jimk/gitwork/parrot/blib/lib
plobsing oh
kid51 Are you talking about something else?
plobsing nope, that would be it. 04:34
kid51 The problem, I think, is that /usr/local/lib/libparrot.dylib is now hard-coded into the binary(ies)
plobsing can you manually copy the lib to there? 04:35
kid51 This looks like the same problem we addressed in trac.parrot.org/parrot/ticket/2005 04:38
plobsing: I tried placing a symlink in that directory. 04:40
Got same results
plobsing the same dyld problem? 04:41
kid51 Yes. But let me try the actual file.
hmm that may have worked 04:42
Reading symbols for shared libraries ......... done
nopaste "kid51" at 192.168.1.3 pasted "backtrace, having copied libparrot.dylib to /usr/local/lib" (15 lines) at nopaste.snit.ch/35587 04:44
plobsing so you aren't able to reproduce the original issue.
kid51 That was at HEAD. I'll try the commit that failed tomorrow.
kid51 must sleep 04:45
04:45 kid51 left
nopaste "bacek" at 192.168.1.3 pasted "Patch for kid51++" (1146 lines) at nopaste.snit.ch/35588 04:51
bacek ouch
too late
04:51 lucian left
dalek rrot: ded4b56 | bacek++ | compilers/opsc/src/Ops/Compiler/Actions.pm:
Insert write barriers into proper places during parsing
 04:53
rrot: 652c6b5 | bacek++ | src/ops/core_ops.c:
Rebootstrap ops with proper write barriers
05:04 mikehh left, mikehh joined 05:40 preflex left 05:42 preflex joined
dalek rrot: a0de3ff | petdance++ | frontend/parrot/main.c:
removed unused Parrot_confess. We have a public one anyway.
 06:04
ttbot Parrot a0de3ff4 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50039 06:06
dalek rrot: cb8b624 | petdance++ | src/interp/inter_misc.c:
removed unused var
 06:20
rrot: ab009a4 | petdance++ | / (2 files):
consted and annotated Parrot_hsh_hash_buffer
sorear Parrot_hsh_ ? Really?
ttbot Parrot ab009a47 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50106 06:22
Parrot ab009a47 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50116 06:23
dalek rrot: 77ee887 | petdance++ | / (2 files):
consting func args in Parrot_pa_count_used()
 06:29
ttbot Parrot 77ee8874 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50192 06:31
Parrot 77ee8874 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50190
dalek rrot: c4a9402 | petdance++ | / (2 files):
allow the ARGIN arg to be ARGIN_NULLOK
 06:37
06:37 fperrad joined
ttbot Parrot c4a94027 i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50250 06:39
06:57 rurban_ joined 06:58 theory left 07:00 rurban left, rurban_ is now known as rurban
dalek rrot: f5a2fed | bacek++ | include/parrot/api.h:
Add foward declaration of Parrot_confess. Fix debug c++ builds.
 07:24
ttbot Parrot f5a2fedf i386-linux-thread-multi make error tt.taptinder.org/cmdinfo/50374 07:27
dalek rrot: 21a351e | bacek++ | include/parrot/api.h:
Add PARROT_EXPORT to Parrot_confess declaration.
 07:41
rrot/opsc_llvm: 085a273 | bacek++ | / (4 files):
Skeleton for LLVM bindings
 08:43
tadzik whoa, whoa 08:57
dalek rrot/opsc_llvm: ed424f8 | bacek++ | / (6 files):
Move global functions loading into LLVM.pm
 08:59
09:56 Teedev joined
dalek rrot: c114262 | jimmy++ | / (4 files):
s/Parrot_hsh_hash_buffer/Parrot_hash_buffer/
 09:56
rrot/opsc_llvm: 632f2d1 | bacek++ | t/library/llvm.t:
Made test to looks more like test.
 10:12
rrot/opsc_llvm: afa93ef | bacek++ | runtime/parrot/library/LLVM.pm:
Add more functions.
rrot/opsc_llvm: 8be7bfe | bacek++ | runtime/parrot/library/ (2 files):
Add Builder
rrot/opsc_llvm: 394c8fe | bacek++ | runtime/parrot/library/LLVM/Module.pm:
Add "private" _get_ptr to fetch underlying pointer.
rrot/opsc_llvm: 4db2e0d | bacek++ | runtime/parrot/library/LLVM.pm:
Add more functions.
rrot/opsc_llvm: 89299c0 | bacek++ | runtime/parrot/library/LLVM.pm:
Add more functions.
rrot/opsc_llvm: 3feabbb | bacek++ | runtime/parrot/library/LLVM.pm:
More functions! More!
rrot/opsc_llvm: 5b2e672 | bacek++ | runtime/parrot/library/LLVM/Type.pm:
Add LLMV::Type
rrot/opsc_llvm: b04aef4 | bacek++ | runtime/parrot/library/Rules.mak:
Add new classes to build.
 10:13
rrot/opsc_llvm: ab2c037 | bacek++ | runtime/parrot/library/llvm_lib.pir:
Include new generated files
rrot/opsc_llvm: eaed6fe | bacek++ | / (2 files):
Add LLVM::Module.add_function
10:14 mikehh left
nopaste "bacek" at 192.168.1.3 pasted "LLVM in Parrot :)" (35 lines) at nopaste.snit.ch/35589 10:16
Tene bacek: Very nice. 10:23
10:48 lucian joined
dalek rrot/opsc_llvm: a3e2d28 | bacek++ | / (3 files):
Add LLVM::Function for real
 10:57
rrot/opsc_llvm: 3a157c6 | bacek++ | runtime/parrot/library/LLVM.pm:
Add more functions.
rrot/opsc_llvm: 9b35f82 | bacek++ | t/library/llvm.t:
Add test for apeend_basic_block.
rrot/opsc_llvm: d473cea | bacek++ | / (5 files):
Add LLVM::BasicBlock
rrot/opsc_llvm: 262226c | bacek++ | runtime/parrot/library/ (4 files):
Add LLVM::Constant
rrot/opsc_llvm: f339d44 | bacek++ | runtime/parrot/library/LLVM/Builder.pm:
Add more guts to Builder.
rrot/opsc_llvm: c906385 | bacek++ | runtime/parrot/library/LLVM/Builder.pm:
Add few actual builders to Builder.
rrot/opsc_llvm: 1ffb6e9 | bacek++ | / (2 files):
Add tests and fix Builder creating
rrot/opsc_llvm: da76a9d | bacek++ | runtime/parrot/library/LLVM/BasicBlock.pm:
Add BasicBlock._get_ptr to use within Builder.
rrot/opsc_llvm: 88d2883 | bacek++ | t/library/llvm.t:
More tests
11:06 Teedev left 11:08 lucian_ joined 11:09 lucian left 11:16 whiteknight joined
whiteknight good morning, #parrot 11:19
tadzik morning 11:27
whiteknight hello tadzik. how are you today?
tadzik oh cool 11:28
11:28 woosley joined
tadzik I'm in the Netherlands, on a hackathon with mberends, masak and jnthn 11:28
whiteknight that sounds awesome 11:29
tadzik aye
Tene What are you guys hacking on? 11:38
tadzik currently it's Yapsi, but it's soon to be a module smoketesting and benchmarking infrastructure 11:39
but don't tell anyone, cos I'm not sure if that isn't a secret project :)
11:39 lucian joined 11:42 lucian_ left 11:48 PacoLinux_ joined 11:51 lucian_ joined 11:54 lucian left, jrt4 left 11:55 lucian joined 11:57 lucian_ left 11:58 lucian_ joined 12:01 PacoLinux_ left, lucian left 12:11 lucian joined 12:12 lucian left 12:15 lucian_ left
dalek rrot/opsc_llvm: e14fb8d | bacek++ | runtime/parrot/library/LLVM.pm:
Fix signature
 12:35
rrot/opsc_llvm: b2c4182 | bacek++ | / (4 files):
Implement Builder.call. Also factor out common functionality to covert
rrot/opsc_llvm: 1076130 | bacek++ | runtime/parrot/library/LLVM/Function.pm:
Implement set_linkage
nopaste "bacek" at 192.168.1.3 pasted "LLVM - work in progress :)" (83 lines) at nopaste.snit.ch/35591 12:36
jnthn bacek: wow! 12:37
bacek whiteknight, look at nopaste :)
jnthn bacek++
whiteknight bacek: nice! 12:38
12:46 PacoLinux_ joined
dalek TT #2040 created by jonathan++: load_ bytecode with a unicode string fails 12:47
TT #2040: trac.parrot.org/parrot/ticket/2040
rrot/opsc_llvm: 6263568 | bacek++ | runtime/parrot/library/LLVM/Type.pm:
Add more types
 12:55
13:00 contingencyplan joined
dalek p/lexical-setting: ebf7adc | jonathan++ | / (6 files):
Merge latest changes from mater into lexical-setting.
 13:01
13:03 PacoLinux_ left
dalek rrot/opsc_llvm: 3ed5117 | bacek++ | t/library/llvm.t:
Create function with args in test
 13:04
13:10 kid51 joined
dalek p: d36ca9c | jonathan++ | / (2 files):
Stub in a --setting= command line argument.
 13:15
p: 93dc13f | jonathan++ | / (3 files):
Re-work the build and bootstrap process to fit the new role of the setting. Since it'll be the outer lexical scope of programs, it needs to be built like the rest of the various bootstrapped bits, not just sa if it's a module.
p: 5be7a2c | jonathan++ | .gitignore:
Update .gitignore.
p: 17d2709 | jonathan++ | / (3 files):
Parsing/compilation of {YOU_ARE_HERE} and inclusion of it at end of setting.
p: 5a4019f | jonathan++ | build/Makefile.in:
Fix make bootstrap-files for setting.
p: fd70004 | jonathan++ | src/stage0/ (4 files):
Update bootstrap to include some of the bits needed for setting support.
p: 0c6b294 | jonathan++ | build/Makefile.in:
Can use --setting=NULL for compiling setting in stage 1 now.
p: b9e1141 | jonathan++ | src/NQP/ (2 files):
Few fixes to the {YOU_ARE_HERE} compilation.
p: b158fac | jonathan++ | src/stage0/ (4 files):
Update bootstrap with compilation fixes.
p: cafdaf5 | jonathan++ | build/gen_setting.pl:
Work around a weird parsing fail - something that shoulda been treated as a block got parsed as a postcircumfix.
p: 3a5e136 | jonathan++ | build/Makefile.in:
Need to fix up various things before we can put Mu and the native types into the lexical setting, so don't do that just yet.
p: 1aa0ec3 | jonathan++ | src/NQP/ (2 files):
First cut of loading NQPCORE.setting by default, overridable at the command line with --setting=NULL or --setting=OTHER. Seems to work, including how it intersects with interactive mode (uses the same outer-setting mechanism...yay). Big missing piece: setting loading not handled in pre-compiled code yet.
p: a32c5f9 | jonathan++ | src/NQP/Compiler.pm:
Make sure if we try to use a setting twice it'll work out.
p: ebf7adc | jonathan++ | / (6 files):
Merge latest changes from mater into lexical-setting.
p: ecaf0c5 | jonathan++ | src/stage0/nqp-setting.nqp:
nqp-setting.nqp (loadable setting) now gone in favour of lexical setting; remove it.
jnthn branch merge heuristic fail!
13:19 PacoLinux_ joined
moritz jnthn: more karma for that jonathan guy :-) 13:24
jnthn :)
That guy commits so much more than me :P
moritz jnthn: it seems that GNU make doesn't like the commented out bits in \\ - continued lines in Makefile 13:25
jnthn moritz: oh, ouch
moritz jnthn: I'll fix it
jnthn moritz: thanks
dalek p: 080d583 | moritz++ | build/Makefile.in:
make GNU make happier
 13:30
13:33 PacoLinux_ left
whiteknight I think the heuristic merge detection looks for at least 15 commits 13:34
tadzik moritz: I have a pull request for you 13:35
13:41 mikehh joined
jnthn Hmm, oddness. I'd have thought IMCC uses declare_lex_preg method on LexInfo to set up lexinfo associations... 13:46
But it seems not to.
jnthn wonders how dependent we are on lexicals living in registers... 13:48
eww...very dependent. 13:55
OK, so, er, IMCC doesn't actually follow the interface that the lexicals spec says it does. :/ 14:00
whiteknight in other news, the world is round, the sky is up, and bears poo in the woods 14:05
jnthn OK, this is designed stupidly. 14:06
In nqpclr I...could just do it rght.
:/
*right
gah, so munch indirection and screwing around when all lexpad incarnation should be is a memcpy :/ 14:07
I can find a way to work around all this I guess.
It may be easier to just do NQP its own lexicals implementation or something, than to try and work within this. 14:08
Ah well, will ponder it over lunch. 14:11
dalek rrot/opsc_llvm: bb2473b | bacek++ | src/dynpmc/ (3 files):
Skeleton for LLVMEngine wrapper
 14:30
rrot/opsc_llvm: af54ffd | bacek++ | runtime/parrot/library/LLVM (2 files):
Fix stupid typo
rrot/opsc_llvm: 699cb37 | bacek++ | runtime/parrot/library/LLVM/Constant.pm:
Add Constant.integer
rrot/opsc_llvm: 9252309 | bacek++ | / (2 files):
Pass return type to add_function
rrot/opsc_llvm: fb705f4 | bacek++ | t/library/llvm.t:
Add 'test' for invoking of LLVM generated function
nopaste "bacek" at 192.168.1.3 pasted "LLVM - intermediate results." (85 lines) at nopaste.snit.ch/35593 14:31
dukeleto bacek: top of the morning, meatbag 14:32
bacek that's it
We can create small function via LLVM, call it and fetch results.
Done. (almost)
And it's 1:30AM...
Good night, humans 14:33
dukeleto bacek++
rurban bacek++ 14:40
whiteknight jnthn: IMCC is far from sacrosanct. If you need to change something there, we're open to hat 14:43
that
14:44 M_o_C joined 14:55 sigue left 14:57 rurban_ joined 14:59 woosley left 15:00 rurban left, rurban_ is now known as rurban 15:02 sigue joined 15:09 ambs joined 15:30 M_o_C left
dukeleto kicks IMCC for funzies 15:30
15:35 sigue left
jnthn I suspect the problem is the entire design of lexical storage so much as IMCC. 15:42
(Upon further reflection.)
15:58 sigue joined
jnthn On the one hand, the model doesn't commit to "lexicals stored in registers". OTOH, the existence of .lex '$foo', $P0 kinda does. 16:11
Coke yawns. 16:12
jnthn o/ Coke 16:13
Coke o/ jnthn 16:15
16:19 Patterner left, Psyche^ joined, Psyche^ is now known as Patterner 17:15 theory joined
whiteknight jnthn: Well, we've known for a long time, months into years, that lexicals are a problem and need a redesign 17:57
jnthn: I think all we are waiting for is somebody to put forward a compelling new design
jnthn whiteknight: Doing NQPLexInfo and NQPLexPad at the moment, to explore.
LexPad is kinda a PMC allocation for no good reason. 17:59
Unless you have a language that needs that. 18:00
whiteknight jnthn: okay, if you come up with something good, we may be eager to incorporate it down into Parrot
jnthn OK.
I think I have a way to keep the current API intact and get what I want and still get it quite cheaply. 18:01
Sadly, LexPad pokes into LexInfo's guts
Without checking it really is a LexInfo 18:02
Segv!
Guess that's fast though :)
whiteknight our new tagline: "Parrot: When we segv, we do it fast" 18:14
tadzik "We don't always segv, but when we do, we do it fast" 18:16
whiteknight the most interesting VM in the world 18:17
dukeleto Parrot: More Coredumps Than You Ever Thought Possible
whiteknight I think IMCC creates the LexInfos, so it can be reasonably certain that the PMC it accesses is a LexInfo 18:34
jnthn In theory IMCC pays attention to the HLL map..
Well, in practice too, thankfully :) 18:36
whiteknight IMCC cares not for your simple theories
Coke seems to recall there's YA branch that needs testing against partcl. 18:43
vtable_overrides, perhaps. 18:44
dalek p/slp: d36a4cd | jonathan++ | / (3 files):
Stub in NQPLexInfo and NQPLexPad, which will eventually give static lexpad support.
 19:10
p/slp: 94fc85c | jonathan++ | src/pmc/nqplexinfo.pmc:
Provide a way to set a static value into a lexpad explicitly, and to indicate when all static values are set.
p/slp: 5c2b972 | jonathan++ | src/pmc/nqplexpad.pmc:
Insert static values into a context's registers, thus putting values from the static lex pad in place per invocation.
19:25 kid51_ joined, kid51 left
Coke partcl "make test" times are noticable slower in a single run of whiteknight's branch. 19:25
I'll run it several times (and then rerun master) to get an average. 19:26
19:26 kid51_ left
whiteknight Coke: okay, thanks 19:27
Coke: does partcl make heavy use of isa, or can? 19:33
because those are the two things which would have gotten slower in the branch 19:34
19:44 kid51 joined 19:54 lucian joined 20:16 bubaflub joined
Coke whiteknight: heavy? I can't say. Certainly it uses those. 20:19
sorear Why are can and find_method separate? 20:21
plobsing because objects can lie? 20:23
sorear it looks like "find_method" overrides are available, and affect can, and affected can even before can overrides were added 20:24
plobsing uncanny 20:26
whiteknight sorear: at the moment, "can" is just a thin wrapper around find_method. VTABLE_can just tells us if find_method returns something 20:31
if that's a semantic relationship we want to keep, that's fine. If we want to be able to decouple the two, that's also fine 20:32
that said, if we absolutely always want VTABLE_can to defer to VTABLE_find_method (and there's an argument to be made for that), we can just rip out the extra vtable 20:33
Coke whiteknight: timings look fine.
whiteknight Coke: fine as in "not noticably worse than master?"
sorear: in Parrot core, there are only two implementations of VTABLE_can: default and Object. Both defer directly to find_method 20:35
and since Object is the mechanism through which all PIR-defined types are written, it is essentially always that way
jnthn I think one of Rakudo's dynpmcs may override it, but it's one that'll go away in the nom branch. 20:38
Oh, maybe it doesn't... 20:39
(as in, doesn't override it)
whiteknight okay, so this is starting to sound like an easy decision
Coke whiteknight: yes. it's pretty even. 20:40
whiteknight Coke: okay. Thanks for checking it. We may not merge that branch in anyway, if "can" is going to be on the chopping block
Coke I think our current vtable override mechanism is painful. Seems like we could do better. 20:41
... which is kind of a non sequitor, I realize.
jnthn I'm quite happy with 6model's way. 20:42
It should be a bunch faster than the way Parrot currently does it.
The idea is probably stealable into Parrot today too, fwiw. 20:44
sorear 6model has vtable overrides now? for how long? 20:46
20:46 lucian_ joined, kid51 left
whiteknight Coke: you're right. The current vtable override system in Parrot is duct tape and paper clips 20:47
blah, VTABLE_get_class isn't overridable either 20:48
jnthn sorear: Didn't finish it up fully yet, but what's left is boring LHF. 20:49
sorear grumbles about jnthn changing 6model while sorear tries to steal it
jnthn sorear: For at least a few weeks, though.
sorear: Note, Parrot vtable override
20:49 lucian left
jnthn They're not very interesting to steal ;) 20:49
I'll give the typecheck cache logic an update worth stealing soon though :) 20:50
sorear Is nqp-nom better than 6model.net at everything yet? 20:51
Coke: you have connections at TPF, right? 20:52
jnthn Not everything, but most.
Well, at least as good at, anyway.
sorear What are nqp/ctmo and nqp/match-nom? 20:53
jnthn match-nom is making Match object use new meta-model 20:56
ctmo = compile time meta-objects
That branch is making me do yak shaving first :)
sorear Is NQP reported by dalek? 20:59
jnthn Yes
I've comitted today and seen the commit messages here and #perl6
Any Parrot folks got any idea why my lexinfo - which works just fine in the .pir file - gets deserialized as a LexInfo? Or appears to... :/ 21:02
plobsing lexinfo deserialized as a lexinfo? I'm not seeing a problem. 21:04
jnthn plobsing: Sorry, i wasn't clear
I have a hll mapped lex info (NQPLexInfo)
It's a dynpmc
When I run the .pir file it's the correct type, which means IMCC musta done the Right Thing and recognized the hll mapping.
However, when I compile it to .pbc, then run that, the type is Parrot's LexInfo 21:05
21:05 fperrad left
jnthn Which is bizzare as my NQPLexInfo isn't even a subclass 21:05
sorear when IMCC is in pbc-mode, are you sure it loads your dynpmc library _at all_?
try adding debugging statements to your VTABLE_init 21:06
jnthn Oh hmm...that is a good question...
sorear it's possible that it never was a NQPLexInfo to begin with
plobsing jnthn: how are you loading the dynpmc library?
jnthn sorear: bingo
plobsing: gist.github.com/857666 21:07
plobsing: turns out the :immediate never gets called in PBC mode.
plobsing yeah. use .loadlib in stead.
whiteknight no, :immediate only gets run by the compiler, not by the loader
welcome to IMCC flag hell 21:08
jnthn whiteknight: Yes, so why doesn't the compiler run it in PBC mode?
plobsing: I'm using .loadlib already
whiteknight jnthn: :immediate is "directly after compilation"
plobsing jnthn: then it is loaded.
jnthn whiteknight: No, it's not.
whiteknight jnthn: :load is "when loaded as a PBC"
jnthn: the implementation would beg to differ
jnthn :immediate means "as soon as IMCC has finished parsing it"
If it didn't, it'd happen too late for IMCC to see the hll mapping and make the correct type of lex info. 21:09
plobsing this argument is moot if .loadlib is already being used. THIS IS NOT THE PROBLEM YOU ARE LOOKING FOR.
whiteknight jnthn: the exact timing with respect to parsing is moot, because :immediate runs during parsing, not PBC loading
jnthn whiteknight: I know!
Apart from IMCC *doesn't* bother to run it during the parse if you use -o x.pbc rather than just running the PIR file right off. 21:10
So the HLL map never takes place
So we create a normal LexInfo instead of an NQPLexInfo
21:10 masak joined
sorear So if you put a "say" in the :immediate sub, then compile to PBC, nothing gets printed by imcc? 21:10
jnthn sorear: Correct, that's how I checked. 21:11
masak hi, I just caused the GC to crash. anyone interested? :)
plobsing dynpmc HLL maps are managed by pmc2c.
whiteknight jnthn: oh, I didn't understand what you are saying. So it doesn't run with -o?
sorear masak: we're busy beating up the PIR compiler. Wanna join?
whiteknight that's definitely weird
jnthn whiteknight: Correct.
masak sorear: not really. :)
jnthn whiteknight: Yes, I think it's weird too
plobsing: Oh...I can do the mapping in the PMC?
plobsing I beleive so. That's how partcl does it IIRC. 21:12
jnthn That may work out better for me. Maybe. :)
sorear It's still a bug if :immediate isn't
jnthn aloha, partcl?
aloha jnthn: partcl is broken, so it's not clear your change breaks it more or github.com/partcl/partcl
jnthn sorear: I agree.
I can make a ticket. 21:13
sorear Can you reproduce the behavior without nqp_group?
jnthn sorear: Almost certainly.
yes, I can 21:14
Filing ticket
dalek TT #2041 created by jonathan++: :immediate ignored when compiling to PBC 21:21
TT #2041: trac.parrot.org/parrot/ticket/2041
jnthn plobsing: OK, I get further with the mapping in the PMC 21:23
Ah, and another fiddle later, things seem to be working out better. 21:25
plobsing++, sorear++
dalek p/slp: 67337e1 | jonathan++ | src/pmc/nqplexinfo.pmc:
Add visit, which should in theory mean all the state of this PMC gets serialized.
 21:26
p/slp: 4f4fa4f | jonathan++ | src/pmc/nqplex (2 files):
Add hll and map directives to NQPLexInfo and NQPLexPad PMCs. Now seems things get serialized as wanted.
21:27 lucian_ left
plobsing jnthn: you know you don't have to explicitly delegate freeze() and thaw() to SUPER right? 21:29
21:30 spinclad joined
jnthn plobsing: oh, duh 21:31
plobsing: Yes, I was at the "anything crazy to make things work" stage :)
21:31 lucian joined
jnthn Will toss, thanks. 21:31
Next piece of fun is realizing that NQP doesn't declare a HLL 21:32
So I can't HLL map anything at the moment!
The yaks. They're lining up all down the street.
plobsing maybe if you let them age, they'll lose their hair 21:33
jnthn :)
21:34 lucian_ joined 21:37 lucian left 21:42 hudnix left 21:45 hudnix joined
cotto ~~ 21:47
masak ::
plobsing __ 21:51
cotto the robot seems to be on another coding rampage 21:58
jnthn (the robot)++
masak (coding rampage)++ 21:59
cotto bacek++
22:04 lucian_ left 22:05 masak left 22:11 masak joined 22:17 masak left 22:20 masak joined 22:24 ambs left 22:25 masak left
NotFound ~~ 22:25
22:29 lucian joined
dalek p: 2e788f6 | jonathan++ | src/ (7 files):
Move NQP into a HLL of its own. This is a pre-req for being able to HLL-map things for NQP's use, but will also go a long way to resolving NQP/NQP-rx symbol conflicts.
 22:54
p: eb5f068 | jonathan++ | src/NQP/ (2 files):
Register NQP as just NQP, not NQP-rx.
p: 4c71035 | jonathan++ | t/hll/0 (2 files):
Update a couple of tests to track NQP moving to its own HLL namespace.
p: 3cc2814 | jonathan++ | src/stage0/ (5 files):
Update bootstrap.
22:57 rurban_ joined 23:00 rurban left, rurban_ is now known as rurban 23:36 lucian left 23:52 bubaflub left
dalek sella: 826d9f4 | Whiteknight++ | src/test/ (6 files):
starting some big cleanups to the test library, specifically refactoring the Loader class. Add docs
 23:55
sella: 8b9db86 | Whiteknight++ | src/test/ (5 files):
refactor the Suite, Result, and Loader classes
sella: b145466 | Whiteknight++ | src/test/Testcase.winxed:
finish test docs and refactors
sella: 93183cb | Whiteknight++ | / (15 files):
improve error message on mockobject verify, listing all un-hit expectations. Improve error reporting for exceptions to show multi-line messages and full backtraces. MockObject.Factory can now return controllers of multiple types. Add tests for tests.
whiteknight let me tell you what: Writing this mock object library is extremely fun 23:57
NotFound winxed -Ofun
whiteknight for serious! 23:58
winxed is turning out to be an extremely pleasant experience
NotFound Warning: redundant option
whiteknight :)
NotFound: I'm thinking the winxed tokenizer logic might make a very nice stand-alone library 23:59

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!